Hosmer-Lemeshow test in SPSS

[Chris Altmayer]

Hello,

I am a new user to SPSS (8.0), and I have a question regarding the Hosmer-
Lemeshow goodness of fit statistic.

When I perform a logistic regression for a specific data set, SPSS
calculates the HL test with only 5 groups (df=3) even though my lowest
cell frequency is only 57.

Why is SPSS limiting me to only 5 groups? The H-L test on only 5 groups is
not very useful. Is there any way to override this?

I have previously down-weighted the data (used analytic weights
divided by design effect) which resulted in the numbers below.
However, can't I still get more groups in the H-L test with the lowest
cell number being only 57?

OUTPUT:

--------Hosmer and Lemeshow Goodness-of-Fit Test--------

BARATALL = No BARATALL = Yes

Group Observed Expected Observed Expected Total

1 2830 2837 64 57 2894
2 2808 2792 85 101 2894
3 2737 2741 156 152 2893
4 2657 2650 236 243 2893
5 1644 1656 363 350 2007

Chi-Square df Significance

Goodness of fit test 4.4164 3 .2199

>Warning # 18585
>The H-L statistic is calculated from fewer than 6 groups.
>Sensitivity to departures from model fit is substantially reduced.

Thank you,
Chris Altmayer
Grad Student